In-Depth Quantitative Multiplex Proteomics Reveal Subtype-Specific Differences Among Functionally Validated AML Stem Cell Populations

Acute Myeloid Leukemia (AML) is a heterogeneous group of myeloid malignancies. The classification of AML subtypes is based on recurrent genetic abnormalities and typical histopathological features, which impact on the patient’s prognosis. AML is also the model disease for the cancer stem cell model with leukemia stem cells (LSCs) residing at the top of a hierarchical organization. LSCs have self-renewal activity and generate leukemic progeny, which make up the majority of leukemic cells. Because LSCs can be quiescent and reside in specific niches in the bone marrow, rendering them resistant to conventional chemotherapy approaches, they are considered the source of relapse. Hence, further strategies to eradicate LSCs are pivotal to improve patient outcomes of this dismal disease.

LSCs present within cell populations can be detected by their capacity to re-initiate the leukemia after xenotransplantation into immuno-compromised mice. However, using current methods, it is neither possible to prospectively isolate pure functional LSCs nor to distinguish them reliably from normal hematopoietic stem cells (HSCs). To define functional LSCs we FACS-sorted primary patient samples of different AML subtypes according to surface-expression of CD34 and CD38 and transplanted each of the resulting four cell populations into conditioned NSG recipients. Several AML samples showed human leukemic engraftment in at least one of the subsets, dissecting LSC-containing and LSC-free subpopulations within the same patient. AML engraftment was mainly observed within the CD34⁺CD38^- fraction, but several cases showed LSC activity also in the CD34⁺CD38⁺ fraction or even in the CD34^- subsets. As age-matched healthy controls, we collected bone marrow hematopoietic stem and progenitor cells (HSPCs, Lineage ^-CD34⁺CD38^-) from individuals older than 60 years, undergoing hip replacement surgery.

In-depth quantitative multiplex proteomic analysis was performed by employing tandem mass tag (TMT) labeling and high-resolution mass spectrometry. Using this approach, more than 7,000 proteins were quantified from 10 LSC-containing and 8 LSC-free fractions from six individual AML samples of different subtypes. Importantly, our data include many low abundance proteins or others that tend to be difficult to detect by mass spectrometry, such as transcription factors and membrane proteins. We identified up to 1500 differentially expressed proteins between LSC-containing and LSC-free fractions. Most interestingly, differentially expressed proteins also clustered according to AML subtype, indicating subtype-specific proteome differences at the level of AML LSCs. Along the same lines, Gene Ontology and Gene Set Enrichment Analyses showed distinct (in some instances even opposing) differences between AML subtypes. For example, interferon- and integrin-signaling were enriched in LSC-fractions of FLT3-ITD, NPM1-mutated AMLs while reduced in LSC-fractions of FLT3-wt, NPM1-wt AML samples. In summary, our data indicate that also at the level of functionally validated LSC populations, subtype-specific differences in protein expression are remarkably evident. This heterogeneity should be taken into account with respect to the development of targeting strategies for LSCs aiming to improve the clinical outcome of AML patients.